code wiki / _hdl_build / nx_v128_share_probe.nx
nx_v128_share_probe.nx
buildroot/runtime/_hdl_build/nx_v128_share_probe.nx
about
nx_v128_share_probe.nx -- F618 rung-1 HOT-LOOP RE-MEASURE (native): what share of the shipped-config
P-frame encode (416x320, emode 673, qp 22 -- the nx_vcodec_layout_gate recipe) is the 16x16 motion
search (vm_search_q, the SAD-dominated kernel the design names first)? Times (a) the full vv_enc_rct9
P encode and (b) an isolated whole-frame vm_search_q sweep with the encoder's own args, same frames.
The sweep is a LOWER BOUND on total SAD-family work (encode also does subpel refine, me8 partition
searches when armed, zero-SAD skip checks, intra SAD). license_tier: ORIGINAL
dependencies 2 imports · 0 importers
imports: nx_syscalls.nxnx_video_codec_wasm.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| 9 | const K_MAGIC_2026: i64 = 2026 |
| 10 | const K_MAGIC_524288: i64 = 524288 |
| 11 | const K_MAGIC_4096: i64 = 4096 |
| 12 | const K_MAGIC_1048576: i64 = 1048576 |
| 13 | const K_MAGIC_8192: i64 = 8192 |
| 14 | const K_MAGIC_1000000: i64 = 1000000 |
| 16 | const SPW: i64 = 416 |
| 17 | const SPH: i64 = 320 |
| 18 | const SPQP: i64 = 22 |
| 19 | const SPEMODE: i64 = 673 |
| 20 | const SPITER: i64 = 30 |
functions
| 22 | func spw(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} sys_write(1,s,n); return 0 } |
| 23 | func spn(v: i64) -> i64 |
| 28 | func sp_fill(f0: *u8, f1: *u8, sz: i64) -> i64 called by 1: main |
| 54 | func main() -> i64 |